Nvidia (NVDA, Financial) saw renewed attention Friday after Morgan Stanley raised its long-term estimates, pointing to a major uptick in demand for inference computing. The firm dismissed the idea that AI is in a "digestion phase" as investor overreaction, calling recent demand trends anything but slow.
Nvidia shares edged up in premarket trading following the report.
Analyst Joseph Moore said that despite broader market concerns around macro trends and supply chains, demand for Nvidia's chips has accelerated, particularly inference GPUs used in artificial intelligence applications. He cited shortages across multiple large language model platforms worldwide as evidence that usage is intensifying.
While Moore held his fiscal 2026 estimates due to export restrictions on the H20 chip, he raised his 2027 forecast significantly. He now expects revenue of $255.5 billion versus a prior $230.9 billion, and adjusted earnings of $6.01 per share, up from $5.37, fueled by ongoing data center strength.
Moore noted that signals from companies like Open Router, as well as internal checks, show inference workloads are rising fast. He said the trend has remained solid even as headlines focus on trade disputes and geopolitical risk — none of which appear to be slowing down actual usage.
He also pointed to a surge in AI commentary across the tech community, with executives from OpenAI, Alphabet (GOOG, Financial), and Grok weighing in on stronger demand signals. These public comments, combined with rising token generation volumes, have reinforced Morgan Stanley's more bullish stance.
What's driving it all? Moore said this isn't hype-driven demand from venture-funded training experiments — it's coming from actual business use cases generating revenue. That, he added, addresses one of the most persistent concerns around Nvidia's stock: that the AI boom might be inflated by speculative capital.
He also pushed back on fears that China's chip stockpiling would dent Nvidia's near-term outlook. “What we're seeing is broad-based growth across regions,” Moore wrote, adding that constrained Blackwell supply and improving Hopper sales point to continued real-world scaling.
